Artificial intelligence
Chatbot ChatGPT will soon be able to remember knowledge about users
ChatGPT should have a memory function. photo
© Frank Rumpenhorst/dpa
So far, the chatbot ChatGPT does not respond to the interests, habits and living conditions of the user. This is about to change soon.
If you then ask ChatGPT to design a birthday card for the child, a jellyfish with a party hat could be in the picture. However, it will take some time before everyone can benefit from the new ability: the function will initially be tested in a small group.
OpenAI wants to take precautionary measures
To ensure that ChatGPT remembers information about users in the future, you can ask the chatbot to do so. At the same time, the software itself can try to pick out knowledge about the user from conversations with it. “ChatGPT’s memory gets better the more you use it,” emphasized OpenAI.
At the same time, the function can trigger new fears that software knows too much about its users. OpenAI wants to take precautionary measures. ChatGPT does not automatically remember sensitive information related to health, for example – but only at the user’s request.
You can also query what the software knows about you – and delete all or individual information. The memory function is intended to make the chatbot more useful. There are temporary chats for conversations without personalization. The information from this will not be used for further training of the software. The memory function can also be switched off completely.
Use in companies
OpenAI also sees benefits for the memory ability when used in companies. This allows the software to remember which format you prefer to receive summaries of meetings at work. Or she could remember the style in which you write your texts and use this when making suggestions for wording.
ChatGPT is the AI chatbot that triggered the hype about artificial intelligence a year ago with expectations ranging from a digital land of milk and honey for everyone to the fear of humanity being wiped out.
AI chatbots like ChatGPT are trained with huge amounts of information and can formulate texts at the linguistic level of a human. The principle behind this is that they estimate, word by word, how a sentence should continue. A disadvantage of this is that the software can sometimes give completely wrong answers, even if it was only based on correct information.